Just noticeable difference in watermarking images

• Main Author: Wong-Jensen, Hans.
• Other Authors: Lin Weisi
• Format: Final Year Project
• Language: English
• Published: 2012
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/48584
• Institution: Nanyang Technological University

In this modern day and age, protecting one’s property is of utmost important as piracy and plagiarism are rampant and prevalent all around us. This is commonly seen over the internet, where digital media is simply used and reused without giving due credit to the owner of that property. Watermarking has been introduced in the past as a form of protection. However, we are able to see that when images or videos are watermarked, it would leave them with an obvious, perhaps annoying, section of the media piece with a logo that the owner has chosen to overlay there. Thus, a watermarking technique that does away with an obvious watermark on an image needs to be developed. The concept of watermarking is not new. However, the various ways in which watermarking can be done are diverse, depending on the usage and embedding of the watermark. In this report, watermarking of an image will be done by fusing the watermark and the image together through the use of Just Noticeable Difference. The resultant product would be an image, exactly, or almost exactly, like the original image, but with a watermark embedded within that the naked eye will not be able to see. In this study, four images will be used and a watermark will be embedded into them by using various different scaling factors to adjust the Just Noticeable Difference thresholds. Once the watermark has been embedded into the images, 2 tests, namely the Peak Signal to Noise Ratio and Structural Similarity Index, will be done to determine how robust and similar the watermarked images are to the original image. Furthermore, the watermark will be extracted from the watermarked images. The similarity test will then be conducted on them to see how close they are to the original ones. After completing the tests on the various images, it has been found that the optimum scaling factor for Just Noticeable Difference in watermarking an image is within the range of 0.5 to 1. Ideally, this scaling factor should be 0.5 as it will not only provide a good watermark recovery rate, but also ensure that the watermarked image is still almost identical to the original non-watermarked image. Essentially, what this report aims to achieve is to create a reliable watermarking scheme that will allow users to copyright and protect their images from being exploited over the internet. By determining ideal Just Noticeable Difference thresholds, any user will be able to watermark his or her images with reliability, i.e., does not compromise the look of the image.